What causes a rectifying diode to fail?

[joecool85]

I fixed the amp.  Put in a new TDA2050 and we're good to go.  One of the traces had lifted from all the heat when the old one blew!   :grr

But it seems fine, I played it for an hour with no issues at all.  Also, they really gooped on the thermal, I think part of the issue may have been too much thermal limiting heat transfer.

[joecool85]

I removed the light bulb limiter info into a new thread since it is so useful.  It's a sticky too.   :tu:

http://www.ssguitar.com/index.php?topic=2093.0

[JPHeisz]

I learned my lesson with diodes (had 1N5406's short and nearly smoked the transformer) - I only use bridge rectifiers now, for anything that draws serious current.
Might be worthwhile to replace those with a 15-25A bridge.

A bridge rectifier is four diodes in a pre-configured package, so it isn't inherently safer.  If you replace lower current diodes with a higher rated bridge rectifier, you might keep the diode/bridge from failing but you'll just be inviting some other problem, since something is still drawing too much current through the rectification circuit. 

Replacing the diodes with a bridge wasn't meant to be 'safer', it's not supposed to act like a fuse. A bridge will withstand a lot more instantaneous current than 4 diodes combined. It keeps on doing what it's meant to do - rectify AC. If there is trouble further down the line, having the rectify blow doesn't help.

[joecool85]

Replacing the diodes with a bridge wasn't meant to be 'safer', it's not supposed to act like a fuse. A bridge will withstand a lot more instantaneous current than 4 diodes combined. It keeps on doing what it's meant to do - rectify AC. If there is trouble further down the line, having the rectify blow doesn't help.

I guess I fail to see the difference between using 4 x 1n4004 diodes or 1 x bridge rectifier.  Electronically they are the same, right?

[JPHeisz]

The spec to look at is Peak Forward Surge Current. 1N4004 has 30A, a 15A bridge will handle ~240A. Sounds excessive but if the power supply has a large capacitor bank to charge (or a fault somewhere), it can come in handy.
I don't fix things, I build them and for anything drawing more than a couple hundred mA, I'll use a bridge.

[Enzo]

The question then arises, are all your other parts also 10x overspec, or just the bridge?

[JPHeisz]

The question then arises, are all your other parts also 10x overspec, or just the bridge?

Would a 1A bridge be 10x overspec for a 200mA supply? Equal to or less than the cost of 4-1N4004. 4 leads to solder, not 8. Higher peak current.
Like I said in my first post, I learned the hard way by nearly losing an expensive transformer through ignorance and trying to save a few cents. To me, it's easy - use a bridge and worry about other things.

[joecool85]

The question then arises, are all your other parts also 10x overspec, or just the bridge?

Would a 1A bridge be 10x overspec for a 200mA supply? Equal to or less than the cost of 4-1N4004. 4 leads to solder, not 8. Higher peak current.
Like I said in my first post, I learned the hard way by nearly losing an expensive transformer through ignorance and trying to save a few cents. To me, it's easy - use a bridge and worry about other things.

I guess if you are building an amp from scratch it would make sense.  But I would think that converting an existing amp to use a bridge would be a pain in the butt.

[Kaz Kylheku]

One thing that puts a load on the diodes in a rectifier is the filter capacitors, during power up.  The capacitors are initially empty, and a surge of current flows through the rectifier to fill them up.  Sometimes you may hear a 60 Hz "twang" when you turn on some power supplies. That's the rush of AC being drawn to fill up the filter caps.

The bigger the filter caps, the harder the full-wave bridge rectifier has to work to load them up on power up. (Some people think that the filter caps are like the amplifier's testicles: the bigger the better).

A leaky capacitor will also add to the workload of the rectifier diodes. Of course any kind of short circuit condition anywhere in the device.

[joecool85]

One thing that puts a load on the diodes in a rectifier is the filter capacitors, during power up.  The capacitors are initially empty, and a surge of current flows through the rectifier to fill them up.  Sometimes you may hear a 60 Hz "twang" when you turn on some power supplies. That's the rush of AC being drawn to fill up the filter caps.

The bigger the filter caps, the harder the full-wave bridge rectifier has to work to load them up on power up. (Some people think that the filter caps are like the amplifier's testicles: the bigger the better).

A leaky capacitor will also add to the workload of the rectifier diodes. Of course any kind of short circuit condition anywhere in the device.

Good point about the PSU caps.  I had wondered about the size of them and how big is necessary.  The LM3886 amps at chipamp.com use PSUs that have 2 x 10,000uF caps!  They're huge!  Of course that same PSU board is used for the stereo and dual mono kits.  For a mono application it's way overkill but for a stereo application it may well be a good idea.

**edit**
Does anyone have a good equation for how large the filter caps should be?

[J M Fahey]

The basic idea is that *all* power supplies have some ripple (unless regulated) which depends, among other things, on capacitance and the amount of current they have to supply.
So you decide "I can live with 4Vpp ripple on my 100W into 4 ohms power amp" and the formula suggests a capacitance value.
If you want half ripple, you double capacitance, just like that.
S the general idea is: double capacitance does not hurt a bit, on the contrary.
On a *tube* rectifier PSU, input capacitors matter and you can't go overboard because the initial current pulse that charges them may (will) stress the tube rectifier, but on SS diodes, that's not an issue.
Economic (or space) considerations will override any technical worries.

[joecool85]

The basic idea is that *all* power supplies have some ripple (unless regulated) which depends, among other things, on capacitance and the amount of current they have to supply.
So you decide "I can live with 4Vpp ripple on my 100W into 4 ohms power amp" and the formula suggests a capacitance value.
If you want half ripple, you double capacitance, just like that.
S the general idea is: double capacitance does not hurt a bit, on the contrary.
On a *tube* rectifier PSU, input capacitors matter and you can't go overboard because the initial current pulse that charges them may (will) stress the tube rectifier, but on SS diodes, that's not an issue.
Economic (or space) considerations will override any technical worries.

So it's not really a concern that 10,000uF caps for instance will burn out the diodes on charge up/turn on?

[J M Fahey]

Not likely.
There is also another unseen current limiting factor there which is transformer copper wire resistance, so in the real world, you will *never* have a zillion Amperes charging current at all.
To put some real world numbers into it, say you have a 22+22VAC transformer with a 30V peak value (a typical chipamp transformer), and internal resistance 1 ohm (actual secondary winding resistance + reflected primary resistance): no matter what size capacitor you will never have more than 30A peak current.
What capacitor size will influence will be for *how long* that charging pulse will last.
We are talking "one time" (turn on) , measured in milliseconds.
A typical diode used there (as part of a bridge) will be a 1N5402 (3 amperes continuous current).
Check its datasheet , http://www.fairchildsemi.com/ds/1N/1N5408.pdf , it's full of very interesting data, that's what they are written for, and you'll see:

Ifsm (non repetitive Peak forward surge current) 8.3 mS (milli Seconds)  single half wave (the duration of the initial 60Hz charging halfwave) = 200A

200 Amperes, it's not a typo.
It means that a typical chipamp power supply,  22+22V transformer rectified with 4 x 1N5402 diodes charging 2 x 5000 or 10000 uF , which I earlier estimated as causing an around 30A peak for a few milliseconds, will never burn the diodes at turn on.
If the chipamp is stereo, it will use a larger transformer, higher capacitance, and a larger bridge rectifier with correspondingly higher single pulse rating, so in practice no sensibly designed PSU will burn "on its own" at turn-on.
There is always a (wide) safety marging built-in.
Then why do power diodes fail now and then?
My guess is that a shorted power amp causes those 30A to flow for a much longer time (although *for us* it still looks quite fast)
, until the fuse actually blows, and *that* stresses them to death.
A fuse may take from 1/4 up to 1 full Second to blow, meaning 250 to 1000 milli Seconds, way over the guaranteed time.
Even so, most times the specified Fuse protects them (not always)

[DartPlayer170]

In power supply design, it is important to calculate the surge current when the filter cap is larger than 1000 uF. A common design is to insert a series surge resistor to limit the surge current rather than using more expensive and larger diodes.

If your TD3050 had a bad solder joint, that could have caused arcing and produced the surge current that destroyed the diode. The shorted diode, in turn, shorted the secondary of the transformer, which blew the fuse.